Hacker News Reader

Comment by jessriedel | Hacker News Reader

Honestly if you could make the training procedure fast and efficient (perhaps even automating that part?), this seems like a very economical litter clean-up technique. Just stick a machine out doors with a solar panel on the top, and come by once a month to empty the bin and refill the food.

You'd need to make sure the exchange rate (food per trash) is set based on the weight of the trash in order to avoid incentivizing the birds from tearing trash into smaller pieces for more food. And you might need some decent machine vision to reject non trash. Seems eminently solveable though.

View on HN

An example of overengineering things.

Animals (specifically corvids) don't measure things by weight and determine an exchange rate, they just look at them with a 1:1 ratio and check for similarity.

In this particular set up, they would simply correlate food with particular types of trash that they have been trained with and dump only those kinds of trash into the apparatus. To them, tearing up the trash would be more effort than finding new trash.

Yes, rather than exchange rate, I'd be more worried about corvids carrying trash from the landfill to exchange it to food. Or, in case of cigarette stubs, emptying well contained ashtrays to the tables and floors only to get to those valuable stubs.

This is more likely actually. Corvids tend to lean towards gaming the system in other ways, like breaking into the trash store and re-dumping the trash repeatedly (like the Indians did with snakes during the British Raj and the Vietnamese did with rats during French rule).

I blame being an operations researcher for always (pessimistically) first and foremost seeing how the system can be gamed. You have to think very very carefully what the objective function is and which kind of undesired solutions need to be forbidden using the constraints.

Guess Corvids are operations researchers then. :)

Once they figure out what the reward system is, they often try to figure out how it can be gamed.

Yes, obviously this is possible. The extent to which it's actually a problem is easy to test. Where I live, it's already necessary to tightly close trash can lids to keep crows and racoons from rooting through it for food.

I'm not suggesting the bird will think in terms of an exchange rate, I'm suggesting they may experiment with random behaviors (like tearing things in half) and it's important that none of those behaviors are re-inforced by leading to more food output. And I do not think it's at all unreasonable to suspect that the bird learns that putting a piece of trash in a slot yields 1 foot pellet, so if they tear the trash in two they can get 2 pellets.

A crow at the beginning of this talk figured out how to bend a metal wire to use as a tool when a straight wire wouldn't work.

https://www.youtube.com/watch?v=0fiAoqwsc9g&t=17s

And the crow in this video figures out how to drink water at the bottom of a bottle by displacing it with pebbles.

https://www.youtube.com/watch?v=xVSr22kqSOs

Ad the water displacement video: It seems to me that the crow would be able to drink the water just fine without throwing the pebble in? Perhaps a cargo cult?

Incidentally one of the next recommended videos after was of my fiancée's cousin who's into studying corvids, and who has been most of my point of reference :).

You assume that to corvids, tearing the trash into two is an easier task to experiment with. It's not. Corvids, just like smarter humans, look for the easiest way to do something. Like another one of the children comments to mine says, they would rather find another way to "reuse" the trash, perhaps by finding some opening to the trash storage, or finding a nearby huge stash of trash which was not the objective, such as a landfill, or experimenting with other materials to see if they work (the reason why mechanical contraptions would barely work with magpies, etc. but would work with most other birds).

You're confidently constructing a very complicated theory of crow psychology here with zero reference to sources except your conversations with family members.

If you're interested, you can always refer to the research from the Ornithology Institute of Oxford University, since that's where they work. Won't give out anymore info, or I'd be doxxing myself.

Related, for things that could be teared up: https://www.theguardian.com/science/2003/jul/03/research.sci....

Apparently, a dolphin Kelly discovered the exchange rate to be one fish per one piece of paper... thus tearing a larger piece of paper into smaller parts, and getting a reward for each small piece. They 'gamed' the exchange rate into their advantage.

> This behaviour is interesting because it shows that Kelly has a sense of the future and delays gratification. She has realised that a big piece of paper gets the same reward as a small piece and so delivers only small pieces to keep the extra food coming.

And dolphins are an exception. Don't they share a very similar brain structure to humans?

Reply on news.ycombinator.com